|Right Channel before changes today|
|Left Channel before changes today|
When I previously looked at R.E.W. last year, it complained that I didn't have a calibrated mike. I did some online research and chose to buy a Dayton EMM-6 calibrated by Cross-Spectrum Labs. With calibration curve, this becomes something like a laboratory quality mike, for just $99. Calibration data files are copied from the USB memory included with the microphone, and then the relevant file is loaded after starting R.E.W, and then it doesn't complain about uncalibrated mike when you press the "Measure" button. So you then have a state-of-the-art audio measuring tool--for $99 instead of the $99,000 it might have cost in 1970. In cases like this, computers are truly wonderful, when powered with incredible free software.
I got the microphone a week or two later, but then the whole project was sitting there for six months until I had to give a long promised and already previously delayed demonstration for a computer club meeting. Unpacking the microphone for the first time, I struggled to get it to work with R.E.W. and my (!@#$%) Windows PC laptop and Emu Tracker Pre. I quickly figured out how to select the Tracker Pre as my audio input device, but still wasn't getting any signal. I was worried that the Tracker Pre might need an AC power adapter, but checking online (on my smartphone at the meeting) it seemed as though it did not need more than USB power except when used standalone without a USB connection. Finally I figured out that this microphone needs phantom power, which a switch on the Tracker Pre turns on.
I measured the living room system with all my existing parametric EQ's operating. The room modes are essentially eliminated already. However there were noticeable defects, and I was immediately motivated to try and fix them. The first was an apparent notch at 80Hz, the crossover frequency, in both channels. I thought to myself, this might be better if I overlap the crossover a bit, perhaps by lowering the Acoustat high pass filter to a lower frequency. I tried 70 Hz, seemed to have no measurable effect except lowering the entire region a tad. So I tried 50Hz, and the notch actually got deeper. So things are not necessarily as they appear in a frequency response graph, and worse, trying to make any correction to a complex speaker/room system is like pushing on string, you never know exactly what is going to happen, and this is especially (though not uniquely) true when you are trying to remove notches.
A picture of the Acoustat response alone (I measured right channel, which also without any crossover shows a series of saw shaped curves below 200 Hz, getting bigger down to a deep notch at 80Hz, followed by a final response peak beginning at 60 Hz and fading below baseline by 40Hz.
So obviously I cannot fix this problem by extending Acoustat response lower, because the problem is in the Acoustat/Room response itself. I didn't try raising the subs higher…that thought just occurred to me writing this. In general I am opposed to using subs higher than necessary. I'm not used to having such a powerful tool give me guidance on such issues.
Leaving that unsolved problem behind, I next looked at a notch around 32 Hz, also in both channels. I had already added a 3dB boost to the left channel, which has a stronger corner loaded position and doesn't seem to have to work as hard to generate high spl's. I made the Q higher and raised the boost to 5db, which seemed to help a little.
Finally I noticed that the response of the left channel was generally rising compared with the right channel, probably helped by that corner position (and also a few eq's selectively added only to the right because of its stronger position). I suppose I could have chosen to lower the sub level. Instead I decided to try a new kind of low frequency tilt filter. I set the center frequency to 20 Hz, a cut of -4dB there, and Q=0.7. That lowers the bass below 100 Hz with a tilt reaching maximum at 20 Hz, where it is 4dB down. (However, by design, only in the subs).
This tilt change measured exactly as expected, and brought right and left channels into much better match, looking reasonably good IMO and still with generous room curve (slightly less OT) below 100 Hz reaching fullness at 20Hz, and fantastic response remaining to 15 Hz and below. (Note for all measurements in this post, the microphone calibration had not been correctly set. This means about 5dB flatter response below 20 Hz, little else of consequence to low frequencies. So the actual response below 20 Hz is better than what is shown here…which is already pretty impressive.)
|Left Channel after changes today|
(Note that both changes were made only to the left sub, so the right sub was same as before.)
Tested bass with Bass Erotica and Grouse We Want to Be Loved. Fantastic bass transparency! However, I decided to roll back the tilt to -3dB to add more casual impressiveness to the bass, a tiny change based more on caution (don't eq too much) than listening test result.
I moved on to the bedroom system, where the Behringer DEQ 2496 had recently died, taking with it the hand tuned room EQ I started developing in about 2005 which reached untouched perfection, or close enough, around 2009. I don't remember where I might have written down the adjustments, it was before this blog started. Things have changed a little anyway, I have new padded hard floor instead of shag carpet, taller bed with head position farther from wall, and other changes. I now know I can program parametric EQ's into the still-working DCX-2496…I just needed to know what adjustments.
Measurement result of full range system with subs was so good I wondered if it even needed any eq.
|Left Bedroom w Sub, No EQ|
|Right Bedroom w Sub, No EQ|
Closer look revealed two defects, which I made some headway correcting. For one, there was a peak around 50 Hz. A similar (and desirable, IMO) peak at 20 Hz (good for room curve) made the 50 Hz peak look innocuous, but it was not--a peak at 50 Hz is far more audible and less desireable. Strangely, I had already dialed in a -5dB cut at 50 Hz (I think a few months ago, I tried doing a quick EQ with pink noise but decided going full range with satellites worked better. My EQ'd version sounded too bass-shy, while the flat out version sounded boomy.)
That 50Hz was in fact the precise place to effect a cut, but it needed a far sharper Q as I had been discovering. So I upped the Q to 4.0, reduced the cut to -4dB, and measured and it looked about as good as possible.
The other defect looks tricky, a response valley centered just above 100 Hz. It may have something to do with the slightly vaulted ceiling and the bed on the floor, a floor-ceiling bounce which results in cancellation around 100 Hz.
Never one to trust stern warnings not to try potentially useful things that might not work, I tried a bit of EQ on this and it seemed to help. I threw a full +4dB at 107 Hz with Q of 3.5. I could possibly throw another +4dB at this without making a bulge, so a "mere" +4dB is a good compromise between reality and ideality.
Bass measurement is now almost fantastic, with a trace of the 100 Hz cancellation remaining, but otherwise reasonably flat, reaching ultimate low peak at about 18 Hz with usable response down to 10 Hz.
Just one day with R.E.W. and everything has changed for the better.